Computers & Industrial Engineering最新文献

{"title":"Research on rail-mounted gantry crane scheduling method of collection and distribution yard for dedicated lane mode","authors":"Houjun Lu ,&nbsp;Peng Ni ,&nbsp;Junjie He ,&nbsp;Minghui Zhang","doi":"10.1016/j.cie.2025.111475","DOIUrl":"10.1016/j.cie.2025.111475","url":null,"abstract":"<div><div>The collection and distribution system is a new solution for addressing the insufficient capacity of container terminal yards. The system features an inland collection and distribution yard with a dedicated transport lane. This paper investigates coordinated scheduling between Rail-Mounted Gantry Cranes (RMGs) and container trucks in a collection and distribution yard under dedicated lane mode. Focusing on the double-layer layout characteristics of the yard, a multi-objective optimization model based on mixed-integer programming is established to coordinate RMGs operations, aiming to simultaneously minimize RMGs completion time while reducing waiting times for both unmanned intelligent container trucks (UICTs) and external container trucks (ECTs). The model rigorously incorporates critical operational constraints including anti-collision requirements, safety distance thresholds for automated double cantilever RMGs (ADCRMGs), and UICTs platooning specifications. Then, a hybrid genetic algorithm-simulated annealing algorithm (GASA) is developed to solve this problem, which integrates the global exploration capabilities of genetic algorithms with the local refinement mechanisms of simulated annealing. Comparative analyses demonstrate the superior performance of GASA over benchmark algorithms including particle swarm optimization and adaptive genetic algorithms. The proposed framework is validated through comprehensive case studies encompassing both small-scale test scenarios and large-scale operational simulations. The proposed model and algorithm are validated through comprehensive case studies based on operational data from Yangshan Port and are found to be highly applicable to the actual operations of the collection and distribution yard.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"209 ","pages":"Article 111475"},"PeriodicalIF":6.5,"publicationDate":"2025-08-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144922942","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From cues to choices: The effect of eco-labels and quality excellence on sustainable online consumption in product–service systems","authors":"Yue Zhu ,&nbsp;Wenyan Song","doi":"10.1016/j.cie.2025.111480","DOIUrl":"10.1016/j.cie.2025.111480","url":null,"abstract":"<div><div>Online shopping platforms create a flexible setting for merging products and services into Product–Service Systems (PSS) that have potential to lessen environmental impacts while promoting sustainable consumption habits. Research remains insufficient in understanding consumer perceptions and evaluations of PSS in digital markets as well as identifying purchase intention drivers. The research uses the Stimulus-Organism-Response (SOR) framework to analyze how environmental cues (sustainable labeling) and functional cues (quality performance) influence consumers’ sustainable online purchase intentions. Our study employs an integrated research design combining Partial Least Squares Structural Equation Modeling (PLS-SEM) with Necessary Condition Analysis (NCA), and fuzzy-set Qualitative Comparative Analysis (fsQCA) to reveal the mediating dual cognitive and affective mechanisms between the variables. Results indicate a dual mediation mechanism in which perceived usefulness (cognition) and perceived enjoyment (affect) jointly transmit the effects of eco-labels and functional performance. The configuration analysis determines various consumer segments through their special combinations of environmental and functional attributes which deliver detailed understanding of diverse online consumption trends. The results provide theoretical insights and practical recommendations for online retailers and policymakers who aim to promote sustainable consumption.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"209 ","pages":"Article 111480"},"PeriodicalIF":6.5,"publicationDate":"2025-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144907878","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Mathematical modeling and optimal control analysis of the monkeypox Clade II infection dynamics under the recent cases of USA","authors":"Mutum Zico Meetei ,&nbsp;Mahmoud H. DarAssi ,&nbsp;Irfan Ahmad ,&nbsp;Muhammad Altaf Khan ,&nbsp;Nurulfiza Mat Isa ,&nbsp;Ebraheem Alzahrani","doi":"10.1016/j.cie.2025.111462","DOIUrl":"10.1016/j.cie.2025.111462","url":null,"abstract":"<div><div>This work presents a mathematical model with optimal control analysis of the monkeypox (mpox) Clade II dynamics, particularly in response to the recent cases reported in the USA. Initially, we develop the model and outline its key properties. We identify the equilibrium points associated with the Clade II model and analyze their asymptotic stability, demonstrating that the Clade II system is locally asymptotically stable (LAS) when the threshold quantity is less than one. We derive the endemic equilibrium point and examine its corresponding results. Considering the reported cases of Clade II in the USA, we employ a nonlinear least-squares curve-fitting method to estimate the parameters. The estimated basic reproduction number, <span><math><msub><mrow><mi>R</mi></mrow><mrow><mn>0</mn></mrow></msub></math></span>, is 3.3282. We conduct a global sensitivity analysis to determine the parameters that most significantly affect the threshold quantity. Furthermore, we formulate an optimal control system for the Clade II virus by introducing controls such as personal protection, treatment of infected individuals, hospital care improvement, and protection from animals (e.g., animal culling). Numerical simulations are carried out for the sensitive parameters. We also simulate both the controlled and uncontrolled systems and discuss the outcomes. The results show that implementing all control measures simultaneously leads to a more significant reduction in the number of infected cases.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"209 ","pages":"Article 111462"},"PeriodicalIF":6.5,"publicationDate":"2025-08-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144866044","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Storage and retrieval sequencing in autonomous vehicle storage and retrieval systems","authors":"Changwen Yu ,&nbsp;Yunfeng Ma ,&nbsp;Yina Hu ,&nbsp;Liang Ren ,&nbsp;Xianyan Yang","doi":"10.1016/j.cie.2025.111468","DOIUrl":"10.1016/j.cie.2025.111468","url":null,"abstract":"<div><div>Autonomous vehicle storage and retrieval systems (AVS/RSs) are pivotal in contemporary logistics, with their performance critically dependent on the coordinated operation of lifts and shuttles. Typically, these systems handle both storage and retrieval requests simultaneously. This paper focuses on the sequencing challenge for dual-command operations conducted by lifts and shuttles in AVS/RSs. We develop a mixed-integer programming model to minimize the makespan of request processing. Additionally, we propose a Branch-and-Bound (B&amp;B) exact algorithm, incorporating an effective lower bound function derived from a relaxed Traveling Salesman Problem model and a pruning strategy based on a dominance rule within a labeling algorithm. Furthermore, we introduce a Beam Search (BS) algorithm enhanced by an Adaptive Large Neighborhood Search (ALNS) mechanism to effectively explore high-quality solutions. Numerical experiments demonstrate that the B&amp;B algorithm consistently yields optimal solutions with significantly reduced computational times compared to the commercial solver Gurobi for small-scale instances. The proposed lower bound and pruning strategy substantially accelerate the convergence speed. For medium- to large-scale instances, the BS-ALNS algorithm outperforms conventional heuristic approaches. Finally, we conduct an analysis of throughput efficiency, revealing that optimizing the shelf tier-to-position ratio and adopting dual-command mode can substantially improve overall operational performance.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"209 ","pages":"Article 111468"},"PeriodicalIF":6.5,"publicationDate":"2025-08-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144866046","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Agent-based collaborative model for forecasting large-scale intermittent spare parts in smart manufacturing industry","authors":"Pan Xiong ,&nbsp;Liping Zhang ,&nbsp;Minzhi Ruan ,&nbsp;Zikai Zhang ,&nbsp;Yingli Li","doi":"10.1016/j.cie.2025.111479","DOIUrl":"10.1016/j.cie.2025.111479","url":null,"abstract":"<div><div>Due to the long lifecycles of large equipment in the smart manufacturing industry, spare parts management has become a significant challenge for automotive and aerospace companies. Though these companies typically stock large quantities of expensive inventory to avoid downtime, the phenomena of “using without preparation” and “preparing without use” also frequently occur. Considering the large-scale intermittent nature of spare parts demand, this study proposes an agent-based collaborative model (ABCM), which integrates a feature extraction agent, a classification agent, and a selection agent to process the large-scale intermittent demand through feature extraction and clustering. The feature extraction agent converts raw demand data into multiple strategies, each characterized by both explicit and implicit attributes, leveraging expert knowledge and autoencoders. The classification agent applies dimensionality reduction techniques to the features derived from a specific strategy, subsequently classifying large-scale spare parts into 7 distinct categories. The selection agent, utilizing a <em>meta</em>-learning approach, effectively associates each category with its best-suited forecasting model, achieving an accuracy of 98.5%. This process ensures a reduction in forecasting errors for each spare part category, thereby enabling more efficient and rapid forecasts of spare part demand. To validate the forecasting performance of ABCM, it was evaluated using datasets from the automotive and aerospace industries. The experimental results demonstrate that ABCM outperforms eight other classical forecasting models and exhibits lower forecasting errors.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"209 ","pages":"Article 111479"},"PeriodicalIF":6.5,"publicationDate":"2025-08-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144889941","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Smart circular economy designed for the decarbonization of waste from industry to promote net zero and sustainable development goals","authors":"Lingfeng Deng ,&nbsp;Riza Sulaiman ,&nbsp;Hanif Baharin ,&nbsp;Ummul Hanan Mohamad ,&nbsp;Bruno A. Pansera ,&nbsp;Domenico Santoro","doi":"10.1016/j.cie.2025.111470","DOIUrl":"10.1016/j.cie.2025.111470","url":null,"abstract":"<div><div>Industrial decarbonization tasks face essential challenges due to the inefficiencies in waste management and operational variations. Traditional methods struggle with real-time adjustments, which are necessary to manage waste volumes and emissions. As a result, these methods are insufficient to meet the Net Zero targets and the Sustainable Development Goals (SDGs)<u>.</u> To address these complexities, the proposed study introduces a Smart Circular Economy (SCE) framework that leverages the advantages of Fuzzy Hybrid Deep Multi-Neural Networks (FH-DMNN) to enhance decision-making and performance in industrial decarbonization. The proposed model utilizes fuzzy logic to manage uncertainty, in conjunction with deep learning, to strengthen waste-to-energy (W2E) conversion, carbon capture, and recycling processes. The simulation of the model is performed under Singapore Industrial CO₂ emission data. It proves its efficacy with 25% improvement in waste processing efficiency and shows its ability in reducing emissions up to 80% by 2060, by supporting SDG 12 and SDG 13. Therefore, the proposed model makes a significant contribution to achieving net-zero emissions in industrial sectors.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"208 ","pages":"Article 111470"},"PeriodicalIF":6.5,"publicationDate":"2025-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144917076","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Agent-based modeling of network-wide metro operations for control strategy evaluation","authors":"Qing Yu ,&nbsp;Qianhui Jiao ,&nbsp;Haoran Zhang ,&nbsp;Xiaolei Liu ,&nbsp;Jiaxing Li ,&nbsp;Jian Yuan ,&nbsp;Xuewu Chen","doi":"10.1016/j.cie.2025.111448","DOIUrl":"10.1016/j.cie.2025.111448","url":null,"abstract":"<div><div>Metro transit has become an essential part of the public transportation system in megacities, where managing network-wide operations poses significant challenges. This research addresses this by developing and applying a novel network-wide Agent-Based Modeling (ABM) simulation platform designed to simulate passengers, trains, and their interactions across the entire metro network, thereby enabling multi-level evaluation of control strategies. Using smart card data from four Chinese megacities (Shanghai, Shenzhen, Nanjing, and Hangzhou), we developed and validated a highly efficient model. Key outcomes demonstrate that the model can simulate a full day’s travel for millions of passengers under 5 min with high accuracy, and that simulating just 3% of passenger flow can sufficiently represent the system’s general performance. Furthermore, tests of various control strategies reveal the existence of ”critical states” in metro operations, where minor supply-side adjustments (e.g., train headways) can avert cascading failures and are more effective than demand-side controls. The primary contributions are: (1) the development of a fast, accurate, and scalable network-wide ABM simulation tool capable of modeling megacity-scale systems; (2) the identification of critical operational states and strategic insights into the effectiveness of different control strategies; and (3) the presentation of a systematic framework for applying such simulation tools to practical policy-making. This work provides a powerful tool for developing more robust and adaptive urban metro transportation systems.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"209 ","pages":"Article 111448"},"PeriodicalIF":6.5,"publicationDate":"2025-08-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144866143","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Analysis of decision policies for two-echelon sustainable supply chain inventory model with learning and forgetting","authors":"Tanmay Halder,&nbsp;Bijoy Krishna Debnath","doi":"10.1016/j.cie.2025.111409","DOIUrl":"10.1016/j.cie.2025.111409","url":null,"abstract":"<div><div>This study analyzed decision-making policies in supply chains facing issues such as imperfect production processes and carbon emissions, considering the worker experience under learning and forgetting. Workers’ engagement in the learn–forget–learn process across various supply chain activities such as machine assembly, operation, inspection, error correction, and material handling enhances both productivity and process quality. To address supply chain challenges, participants shift from individual competition to collaborative strategies for long-term sustainability and productivity. The primary focus of supply chain participants is on sustainability, with a particular emphasis on improving worker experience to drive productivity and process quality. In examining decision-making approaches, three policies emerge: Policy-I advocates for distributing decision-making power equally among supply chain members, such as manufacturers and retailers. Policy-II designates the manufacturer as the leader, with the retailer following suit. Conversely, Policy-III positions the retailer as the leader and the manufacturer as the follower. In this paper, the production rate is first approximated by incorporating human factors such as skill, fatigue, physical ability, and motivation, with the learning-forgetting effect. Optimization problems are then formulated under three policies, considering both with and without learning-forgetting effects. The models are solved using Genetic Algorithm (GA) and Particle Swarm Optimization (PSO). A statistical analysis is performed to compare their performance, with parameter optimization carried out using the Taguchi method. A comparative study of supply chain performance under three decision-making policies reveals the significant impact of learning-forgetting effects on decision-making and performance outcomes. This enables managers to identify the most effective strategy for optimizing supply chain operations, ensuring the selected policy aligns with organizational goals and adapts to the complexities of real-world supply chain systems. Finally, the sensitivity analysis shows how changes in particular parameters influence supply chain performance.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"209 ","pages":"Article 111409"},"PeriodicalIF":6.5,"publicationDate":"2025-08-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144904096","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evolutionary experience-guided deep reinforcement learning for job shop scheduling with AGV transportation constraints in flexible manufacturing system","authors":"Chen Li ,&nbsp;Lin Lin ,&nbsp;Xuze Qiu ,&nbsp;Xiyan Zhao ,&nbsp;Wenqiang Zhang ,&nbsp;Mitsuo Gen","doi":"10.1016/j.cie.2025.111465","DOIUrl":"10.1016/j.cie.2025.111465","url":null,"abstract":"<div><div>Flexible job shop scheduling with transportation constraints (FJSPT) is crucial in intelligent manufacturing, especially with the widespread adoption of automated guided vehicles (AGVs), which significantly impact production efficiency. However, existing research often simplifies transportation modeling, leading to scheduling strategies misaligned with real-world scenarios. FJSPT involves complex dependencies among operation sequencing, machine assignment, and AGV scheduling, making it highly challenging. Traditional methods often suffer from local optima or high computational cost, limiting scalability. In contrast, deep reinforcement learning (DRL) offers stronger adaptability and global optimization potential. To improve DRL’s effectiveness for FJSPT, this study proposes a graph transformer-based genetic algorithm-guided DRL (GT-GADRL) algorithm. First, a heterogeneous graph is constructed to capture processing and transport dependencies among operations, machines, and AGVs. Secondly, a graph transformer-based mechanism integrates local feature extraction and global information fusion to enable efficient representation learning and composite scheduling, enhancing generalization and decision-making efficiency. Additionally, evolutionary experience-guided imitation learning strategy is introduced to accelerate training convergence with expert demonstrations. Furthermore, a baseline-based REINFORCE algorithm further optimizes the policy network, overcoming expert limitations and enhancing exploration capabilities. Experimental results on diverse FJSPT benchmarks show that GT-GADRL consistently outperforms existing methods in solution quality and training efficiency, demonstrating strong robustness, generalization, and adaptability.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"209 ","pages":"Article 111465"},"PeriodicalIF":6.5,"publicationDate":"2025-08-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144878790","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Hierarchical multi-agent cooperative distributed fault diagnosis method for large-scale complex industrial plant with deep reinforcement learning","authors":"Yan Liu ,&nbsp;Zuhua Xu ,&nbsp;Jun Zhao ,&nbsp;Chunyue Song ,&nbsp;Zhijing He ,&nbsp;Kai Wang","doi":"10.1016/j.cie.2025.111463","DOIUrl":"10.1016/j.cie.2025.111463","url":null,"abstract":"<div><div>Large-scale industrial plants are characterized by multi-unit interaction and large quantity of process variables, causing unsatisfactory performance of conventional centralized fault diagnosis methods. To address it, a hierarchical multi-agent cooperative distributed fault diagnosis method with deep reinforcement learning is proposed for complex industrial plants. Firstly, we developed a hierarchical multi-agent distributed fault diagnosis model, in which two types of agents are designed in hierarchical framework: one is intra-unit agent, which is utilized to adaptively learn the distributed model construction policies; the other is inter-unit agent, used to explore the relational dependencies between different units. In such hierarchical multi-agent cooperation framework, the dimensionality curse and the multi-unit coupling issues of the centralized fault diagnosis can be handled. Based on the decision of these two agents, we develop a sparse hybrid kernel matrix to effectively transfer the meaningful modeling information of related units and discard the redundant information of unrelated units for disturbed fault diagnosis enhancement. Then, the training of above distributed diagnosis model is transformed into a multi-agent cooperation decision problem in deep reinforcement learning. To solve it, a hierarchical multi-agent deep deterministic policy gradient training strategy is proposed to iteratively train the intra- and inter-unit agents through real-time interaction with the distributed fault diagnosis model, and the prior physical knowledge is embedded as learning regularization to steer state encoding and the agents’ actions to adhere to underlying physics. Finally, the effectiveness of the proposed method is demonstrated on two real-world large-scale industrial plant cases. The results have demonstrated that the proposed method can achieve an average fault diagnosis rate of 94% and average false alarm rate of 0.34% in complex industrial scenarios, surpassing existing similar methods in most average metrics. These findings offer novel perspectives on the application of deep reinforcement learning in industrial fault diagnosis.</div></div>","PeriodicalId":55220,"journal":{"name":"Computers & Industrial Engineering","volume":"209 ","pages":"Article 111463"},"PeriodicalIF":6.5,"publicationDate":"2025-08-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144866142","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}